nutanc
Re:Astra can read and write weights to improve Qwen
on Tuesday November 03, 2026 @06:53AM (#50350854)
I don't think weight space interpretability is the only thing. I think it's the fact that Astra(and apparently Opus 5.5) can directly read the weights and find patterns and make changes. Writing code for interpretability is one thing. I think this approach shows models now understand weights directly without tools.
Re:Opus 5.5 is good at explainer videos
on Friday September 25, 2026 @08:43AM (#49843857)
Would this means remotion etc are not needed?
Re:Speech Recognition and TTS in less than 500kb
on Sunday July 19, 2026 @12:41AM (#48965039)
This is awesome. I am trying to build a full scale ASR system within 20-25MB. Now that we have Claude code to run experiments, I have started running some experiments. Promising results so far. First realization is that you can capture the nuances of speech in just 3300 embedding vectors(786d). This sequence can be decoded with a small CTC system to get text. Next experiments are on reducing the 768 dimension space into a 64D space. Thats also show some promising results. Hooking up my system so that the agent blogs the results everyday[1]. So my research "claw" setup does the experiments and posts results which I check in the morning and adjust the experiment direction as needed. Its not fully automated yet, but almost there.
[1] https://blog.trulm.com/posts/speech-as-independent-parts/
Ask HN: If AI existed before CRMs, what would we have?
People say if computers were invented before cars we would have only had self driving cars. So if AI existed before CRMs, what would we have?
Re:
on Tuesday May 19, 2026 @02:07PM (#48196937)
You can get very close to a working version by reverse engineering the javascript with Claude. I got a good version almost an year back. With Opus it might do a better job.
Re:OpenAI's WebRTC Problem
on Saturday May 09, 2026 @01:14AM (#48072015)
Most of the problems happen because we want to simulate human conversations. While thats a good goal to have, another approach is to let the user know clearly they are talking to a bot. You will be surprised at how accomodating users can be when they know they are talking to a bot and want their queries resolved.
Killing SaaS part 2. What's the playbook for distributing your product
Killing SaaS. Anatomy of a murder. How I replaced Wisprflow.ai with vibe coding
Template for telling the truth on layoffs like at Block
For the people who have been laid off, I think the letter by Jack is a disservice. So I crafted a more direct letter. Hope future letters follow this approach.Today we are making one of the hardest decisions in the history of our company: we are reducing our organization by nearly half, from over 10,000 people to just under 6,000. This means over 4,000 of you will be leaving the company.I want...
The most important part of an AI system-the human
Ask HN: Built an interactive textbook for math. Non profit or startup route?
We have created a product which is an interactive textbook for math. It covers the CBSE curriculum in India. The idea being that in this day and age why should students read from static textbooks and digital textbooks should mean much more than PDF links. Now that we have all the content we have to make a choice. Should we go the Khan Academy route and become a non profit(Highest impact). Or go...
Re:AI Is Killing B2B SaaS
on Thursday February 05, 2026 @05:29AM (#46898100)
Lets break this down. There is very little in newness in what Anthropic announced. Claude had skills for a long time. They have added one more layer of abstraction and called it plugins. This mainly comes with a set of integrations.
Thats the pitch.
But, what are Claude plugins?
Plugins=Commands+Skills+Integrations.
Commands are specific to Claude code. But commands and skills are nothing but prompts at their basest level.
So what is the main differentiator?
Integrations.
But what are you integrating with?
SaaS companies.
And what is the stock market doing?
Dumping SaaS stocks.
How do they think Claude cowork will work without the integrations. Without the system of records.
If anything, these SaaS products have become more important. If I was a trading guy, I would go to the github of claude plugins, see the default integrations and buy the stock of those companies.
Re:Claude Cowork and the Case of SaaSpocalypse
on Wednesday February 04, 2026 @11:58PM (#46895840)
Claude cowork and the SaaSpocalypse case makes no sense.
What are Claude plugins?
Plugins=Commands+Skills+Integrations.
Commands and skills=prompts
So differentiator?
Integrations.
Integrating with?
SaaS companies.
And what is the stock market doing?
Dumping SaaS stocks.
Claude Cowork and the Case of SaaSpocalypse
Re:A copyleft draft license for datasets. You take my data, you give y...
on Wednesday February 04, 2026 @12:26AM (#46881825)
2 years since this started. The effort is to protect the creators from the AI machine.
The AI machine is different from the printing press. We think we need protections more than just the copyright. Can we have a copyleft for data?
In this age of AI, how do we protect the creators?
A copyleft draft license for datasets. You take my data, you give your data
Part 1: IndiaAI mission does not need compute, it needs data
Show HN: Learning Path for CBSE Math
https://books.innings2.com/learningpath
Math is a hierarchical subject. You build up your concepts one by one. So we have created an interactive learning path for CBSE math. Just click on any chapter and see the pre requisites for that chapter. Some chapters will need lots of pre requisites. Some need a lot less. Want to refresh your learning, just double click and do the interactive text book :)
Re:AI friend- Brought to you by your friendly neighborhood mega corpor...
on Saturday January 17, 2026 @02:21AM (#46656012)
Chat GPT has ads now. This is a dangerous domain we are entering and society is not ready.
AI friend- Brought to you by your friendly neighborhood mega corporation
Show HN: A diagnostic report for students studying math
https://books.innings2.com/pages/cbse_math_worksheets.html
Now that the 10th grade CBSE exams are coming close, was trying to see how can edtech and AI help the students. Our goal was not exam tech as there are enough startups around it. So we decided on diagnostics. Can a student do a simple worksheet and as a result we give an AI powered diagnostic report showing where the kid is lagging and what "concepts" they have misunderstood? The result is our...
The Ghost in the Machine: How I learned to stop worrying and love the AI
Re:The Ghost in the Machine: How I learned to stop worrying and love t...
on Friday January 02, 2026 @12:34AM (#46461756)
Organizations adapting AI is the biggest problem that businesses are facing right now. Even in Ozonetel I face this problem day in and day out. The employees who really use AI to its full potential are minuscule. I can count on my fingertips. We need to overcome this in the right way or we will face the same problems we faced during industrial revolution.
Re:ARR is dead. Long live VRR: Vibe Revenue Run-rate
on Tuesday December 30, 2025 @12:09AM (#46429742)
Got pissed off with too many ARR manipulations and AI startups announcing revenue numbers manipulations(best day * 365 as ARR etc). These startups are not only messing up the AI ecosystem with the non standard numbers, they are also messing up the SaaS ecosystem by co opting the SaaS metrics. Now SaaS startups are supposed to show the same scale though the AI startups have also not achieved that scale. So here is what I propose, the AI startups should use their own new vocabulary. Since everyone is vibe coding, I suggest VRR, Vibe Revenue Run-rate :)
I have even provided a formula(all scientific and all) and also provided a checklist for the VCs.
ARR is dead. Long live VRR: Vibe Revenue Run-rate
Open Source AI Reclaims the Digital Commons
Re:Marx's Theory of the "Commons" Explains the AI Crisis
on Monday December 22, 2025 @12:51AM (#46351723)
You can disagree with the commons definition, and thats fine. But the point I wanted to make was about the exploitation. The open Internet was built with a code of sharing. Now they are trying to put walled gardens around all that knowledge. Lets remove Marx from the equation if that becomes a bone of contention. But we as a society need to come up with better dialogues to decide how we will treat our creators and how we will deal with the AI copy machine. We cannot expect that profit mongers will do the right thing.
Marx's Theory of the "Commons" Explains the AI Crisis
Why is human downloading a file called pirating and AI scraping called training
In this age of AI, with respect to copyright, looks like AI has more freedoms than humans. Sites like Anna's Archive, the PirateBay etc are blocked for humans(in India for example) and if you download and read a book, its called piracy. But if the same book is fed to an AI for training, apparently its fine and dandy. So Artificial Intelligence has more freedoms than Actual Intelligence?